|Ph.D Thesis||Department of Electrical Engineering|
|Supervisor:||Distinguished Prof. Shamai )Shitz( Shlomo|
The traditional concept of the seven layer OSI model considers each layer separately with no coupling between layers, leading to separate algorithms and optimizations for each layer. Indeed this traditional concept is suitable for wireline systems where the physical layer
reliability is high and transmission failures are caused by other reasons such as collisions. However, in wireless environment the fading causes long bursts of transmission errors which contradicts the previously mentioned reliability assumption. Thus classic network protocols which disregard the fading realization perform poorly in fading environment. On the other hand, by the known water-filling result of information theory, the maximal throughput is attained for
infinite codeword disregarding any delay criterion. Thus new type of algorithms which combine the physical layer and the networking layer are required.
In this work we propose two algorithms analyzed using the Shannon theory capacity expressions. The first algorithm controls the tradeoff between throughput vs. delay in multi user environment by using a new scheduling concept which consider both fairness and fading conditions. The algorithm can implement TDMA, the maximal throughput achieving
policy, or any combination of the two by setting user controlled parameters. The second algorithm considers a single user model in which the optimal power/rate policy is found for minimal average delay under average power constraint. The optimal power/rate policy is found
by defining a new optimization criterion combining power and delay, and solved by using the dynamic programming algorithm. This concept can be used for other optimization criterion such as minimal buffer overflow probability as well as extensions to multi user environments.